Energy-efficiency maximization for OFDMA networks with total power constraint

This paper addresses an energy-efficiency (EE) maximized resource allocation (RA) problem for a downlink orthogonal frequency division multiple-access (OFDMA) network, when the system's total power (including transmission power and fixed circuit power) is constrained below a peak value. The EE is evaluated by the number of message bits delivered corresponding to per Joule energy consumed by the whole system. Instrumental to the idea behind the RA algorithm design is to first find the optimum total power leading to the maximum EE, and then the spectral efficiency (SE) maximized RA corresponding to that optimum power is the optimum RA to maximize the EE. This idea is novel and significantly different from those used in previous works. Specifically, the maximum EE when the BS consumes a given total power ρ, denoted by η(ρ), is derived, and theoretical analysis is made to study properties of η(ρ) which leads to much simplicity and insight for the design of an EE-maximized RA algorithm. Most interestingly, intuitive yet insightful geometric interpretation is presented to corroborate that analysis. Furthermore, the analysis and geometric interpretation reveal important insight that the SE-maximized RA leads to the maximum EE when the peak total power is small, while it leads to a significantly smaller EE than the EE-maximized RA when the peak total power is very high. Moreover, that analysis also leads to much elegance for the RA algorithm design, in that motivated by the properties of η(ρ), the bisection method can be used in combination with the sensitivity analysis in convex optimization theory to design a low-complexity yet effective RA algorithm in a very simple way. The theoretical analysis and effectiveness of the RA algorithm are corroborated by numerical experiments.